Multiple feature temporal models for the semantic characterization of video contents